Climate and hydrologic ensembling lead to differing streamflow and sediment yield predictions

Climate change is leading to alterations of the hydrologic cycle and sediment movement within watersheds, but the details and impacts of these changes are indeterminate. To reduce this uncertainty, many researchers create ensembles by averaging the projected temperature and precipitation from multiple global climate model (GCM) ensemble members before running these as forcing inputs through hydrologic models. There is little research quantifying if these ensembled climate scenarios produce similar hydrologic model results to those based on individual ensemble members. We created multiple sets of ensembled climate inputs for a pair of hydrologic and sediment yield models of adjacent watersheds that drain to the Great Lakes. We then compared the hydrologic and sediment results of the models forced by these ensembled climate scenarios with hydrologic ensembles created by running the individual climate ensemble members through the same hydrologic models. We found that, in all cases, the streamflow and sediment yield results are significantly different at the 5% confidence level and the ensembled climate scenarios can lead to systematic negative biases. We also looked at three subset hydrologic ensembles: all 10 CMIP5 ensemble members from the CSIRO mk3.6 model; a Representative ensemble with high, moderate, and low precipitation predictions; and a Best Fit ensemble based on GCM performance relative to historic climate. We found that the subset ensembles covered a large portion of the range of outputs for the whole set, while producing mean annual streamflows within 5.5% of the full hydrologic ensemble results and sediment yield and sediment discharge results within 12.2%.